GPU-based Parallel Differential Evolution with Local Pattern Search on Function Optimization

摘要

This paper presents a parallel Differential Evolution (DE) algorithm with local search for function optimization problems, with graphics hardware acceleration. Differential Evolution is population-based meta-heuristic, originally designed for continuous function optimization. Graphics Processing Units (GPU) is an emerging desktop parallel computing technology that is getting popular with its widespread adoption. In this paper, the classical DE is implemented in a GPU platform with CUDA? technology and a local Pattern Search is added to enhance its search ability. The test results show great savings in computation times and demonstrate a promising direction for high speed optimization on a desktop computing setting.